Error bars - which kind, and saying which

1 · Learn the move · Cognitive verifier

An error bar is a claim, and three different claims wear the same whiskers. Standard deviation describes the spread of your observations. Standard error describes the precision of the mean - and shrinks as n grows, which is why it flatters small differences. A confidence interval makes the precision claim at a stated coverage. A bar without a label is unfalsifiable decoration, and reviewers know it. The cognitive verifier move: before any bar is drawn, the model walks three sub-questions with you - what is n, and is it biological or technical? What claim is this figure making, spread or precision? Which interval expresses that claim? The model may not pick; unanswered questions become [NEEDS] lines, and the caption states the kind and the n or the figure does not ship.

Before drawing error bars on [figure], verify by sub-questions, one at a time: (1) What is n per point, and is it biological or technical replication? (2) Is the figure's claim about the spread of observations or the precision of the mean? (3) Given those answers, which bar expresses the claim: SD, SEM, or a CI at stated coverage - and what would each alternative wrongly imply here? Do not choose for me. Then draft the caption sentence stating the bar type and n, with [NEEDS: ...] for anything unanswered.